linux-kernel - Re: [V8 1/2] irqchip: gic: Add support for multiple MSI for ARM64

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Tue, 23 Sep 2014 10:04:37 -0700
From:	Suravee Suthikulpanit <Suravee.Suthikulpanit@....com>
To:	Thomas Gleixner <tglx@...utronix.de>
CC:	<marc.zyngier@....com>, Mark Rutland <mark.rutland@....com>,
	<jason@...edaemon.net>, <pawel.moll@....com>,
	<Catalin.Marinas@....com>, <Will.Deacon@....com>,
	<liviu.dudau@....com>, <Harish.Kasiviswanathan@....com>,
	<linux-arm-kernel@...ts.infradead.org>,
	<linux-pci@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
	<linux-doc@...r.kernel.org>, <devicetree@...r.kernel.org>
Subject: Re: [V8 1/2] irqchip: gic: Add support for multiple MSI for ARM64

Thomas,

Sorry again for the mistake on my part. Let me try to address some other 
concerns you have below.

On 09/22/2014 04:08 PM, Thomas Gleixner wrote:
> On Sat, 20 Sep 2014, suravee.suthikulpanit@....com wrote:
>
>> From: Suravee Suthikulpanit <Suravee.Suthikulpanit@....com>
>>
>> This patch implelments the ARM64 version of arch_setup_msi_irqs(),
>> which does not return 1 for when PCI_CAP_ID_MSI and nvec > 1.
>
> I can see that myself. What your changelog is missing is the reason
> WHY you think that copying that code from drivers/pci/msi.c and
> removing the "PCI_CAP_ID_MSI and nvec > 1" has any value.

[Suravee] This is mainly be cause the weak version of 
arch_setup_msi_irqs() in the drivers/pci/msi.c doesn't support 
multi-MSI. Sorry for not being clear in the commit message.

>
> And that new function "arm64_setup_msi_irqs" is declared in which
> header file exactly?

[Suravee] This was supposed to be arch_setup_msi_irqs(). My bad. I'm 
fixing this in the next version.

......

>> + *
>> + * Note:
>> + * Current implementation assumes that all interrupt controller used in
>> + * ARM64 architecture _MUST_ supports multi-MSI.
>
> Great assumption....
>

[Suravee] So, Marc and I have discussed in the past that at this point, 
we are not seeing the case that there will be interrupt or 
MSI-controller that will not support multi-MSI.  If you think this 
should not be the case, would you please share your thought.

......

>
> At least you are consistent on the useless side of affairs:
>
>> +{
>> +	struct msi_desc *entry;
>> +	int ret;
>> +
>> +	list_for_each_entry(entry, &dev->msi_list, list) {
>> +		ret = arch_setup_msi_irq(dev, entry);
>
> Anyone who has the slightest idea how multi-MSI works will know that
> this CANNOT work at all, but that's none of my business.

[Suravee] I noticed that in the x86 version, there is a callback that 
each MSI controller need to register for handling the multi-MSI stuff.

In gicv2m_setup_msi_irq(), there is logic which handles the setup for 
multi-MSI and MSIx separately. In case of multi-MSI, the vectors are 
allocated on the first call to arch_setup_msi_irq(). Here, Marc and I 
are trying to simplify the arch-specific code so that each GIC 
controller (V2m and V3) would not need to implement and register the 
callbacks separately for handling multi-MSI.

The thing that is broken here is the error handling where the 
arch_setup_msi_irqs() is supposed to return the number of available MSI 
vectors. It would fail to do so because the arch_setup_msi_irq() would 
not return positive value. We should be able to fix this by 
re-implementing the arch_setup_msi_irq() and arch_setup_msi_irqs() to 
allow returning of positive values.

Please let me know what you think. I am open for suggestions :)

Thanks,

Suravee
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/